Holomorphic Morse inequalities and
the Green-GrifRths-Lang conjecture
Jean-Pierre Demailly
Universite de Grenoble I, Departement de Mathematiques
Institut Fourier, 38402 Saint-Martin d'Heres, France
e-mail: demaillyOf ourier . ujf -grenoble . f r
Dedicated to the memory of Eckart Viehweg
Abstract. The goal of this work is to study the existence and properties of non
entire curves / : C — )■ X drawn in a complex irreducible n-dimensional variety X ,
and more specifically to show that they must satisfy certain global algebraic or
differential equations as soon as X is projective of general type. By means of
holomorphic Morse inequalities and a probabilistic analysis of the cohomology of
jet spaces, we are able to prove a significant step of a generalized version of the
Green- Griffiths-Lang conjecture on the algebraic degeneracy of entire curves.
Resume. Le but de ce travail est d'etudier I'existence et les proprietes des
courbes entieres / : C — )■ X tracees sur une varete complexe irreductible de
dimension n, et plus precisement de montrer que ces courbes doivent satisfaire a
certaines equations algebriques ou differentielles globales des que X est projective
de type general. Au moyen des inegalites de Morse holomorphes et d'une analyse
probabiliste de la cohomologie des espaces de jets, nous demontrons une premiere
etape significative d'une version generalisee de la conjecture de Green-Griffiths-
Lang sur la degenerescence algebrique des courbes entieres.
Key words. Chern curvature, holomorphic Morse inequality, jet bundle, co-
homology group, entire curve, algebraic degeneration, weighted projective space,
Green-GrifRths-Lang conjecture
Mots-cles. Courbure de Chern, inegalite de Morse holomorphe, fibre de jets,
groupe de cohomologie, courbe entiere, degenerescence algebrique, espace projectif
a poids, conjecture de Green-Griffiths-Lang.
MSG 2010 Classification. 32Q45, 32L20, 14C30
0. Introduction
Let A be a complex n-dimensional manifold ; most of the time we will assume
that A is compact and even projective algebraic. By an "entire curve" we always
mean a non constant holomorphic map defined on the whole complex line C, and
we say that it is algebraically degenerate if its image is contained in a proper
algebraic subvariety of the ambient variety. If : A — )■ A is a modification and
2
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
/ : C — )■ X is an entire curve whose image /(C) is not contained in the image
of the exceptional locus, then / admits a unique lifting / : C — ?■ X. For this reason,
the study of the algebraic degeneration of / is a birationally invariant problem,
and singularities do not play an essential role at this stage. We will therefore
assume that X is non singular, possibly after performing a suitable composition
of blow-ups. We are interested more generally in the situation where the tangent
bundle Tx is equipped with a linear subspace V C Tx, that is, an irreducible
complex analytic subset of the total space of Tx such that
(0.1) all fibers Vx :=V f] Tx,x are vector subspaces of Tx,x-
Then the problem is to study entire curves / : C — > X which are tangent to V,
i.e. such that /*Tc C V . We will refer to a pair (X, V) as being a directed variety
(or directed manifold). A morphism of directed varieties $ : {X,V) — ?> (y, VF) is
a holomorphic map $ : X — )■ y such that C W; by the irreducibility, it
is enough to check this condition over the dense open subset X \ Sing(l/) where
V is actually a subbundle. Here Sing(F) denotes the indeterminacy set of the
associated meromorphic map a : X ---> Gr{Tx) to the Grassmannian bbundle of
r-planes in Tx, r = rankF ; we thus have V|X\Sing(T/) = oi* S where S — )■ Gr(Tx)
is the tautological subbundle of Gr{Tx)- In that way, we get a category, and we
will be mostly interested in the subcategory whose objects (X, V) are projective
algebraic manifolds equipped with algebraic linear subspaces. Notice that an entire
curve / : C — >■ X tangent to V is just a morphism / : (C, Tc) — >■ (X, V).
The case where V = Tx/s is the relative tangent space of some fibration
X — )■ 5" is of special interest, and so is the case of a foliated variety (this is the
situation where the sheaf of sections 0{V) satisfies the Frobenius intcgrability
condition [0(V),0(F)] C 0{V)); however, it is very useful to allow as well non
integrable linear subspaces V. We refer toV = Tx as being the absolute case. Our
main target is the following deep conjecture concerning the algebraic degeneracy
of entire curves, which generalizes similar statements made in [GG79] (see also
[Lang86, Lang87]).
(0.2) Generalized Green-GrifRths-Lang conjecture. Let (X, V) be a projec-
tive directed manifold such that the canonical sheaf Ky is big [in the absolute case
V = Tx , this means that X is a variety of general type, and in the relative case
we will say that (X, V) is of general type) . Then there should exist an algebraic
subvariety Y C. X such that every non constant entire curve / : C — >■ X tangent
to V is contained in Y .
The precise meaning of Ky and of its bigness will be explained below - our
definition does not coincide with other frequently used definitions and is in our view
better suited to the study of entire curves of (X, V). One says that (X, V) is Brody-
hyperbolic when there are no entire curves tangent to V. According to (generalized
versions of) conjectures of Kobayashi [Kob70, Kob76] the hyperbolicity of (X, V)
should imply that Ky is big, and even possibly ample, in a suitable sense. It would
then follow from conjecture (0.2) that (X, V) is hyperbolic if and only if for every
0. Introduction 3
irreducible variety Y C X, the linear subspace V~ = T~^^n fi^^V C T~ has
a big canonical sheaf whenever /j, : Y ^ Y is a, desingularization and E is the
exceptional locus.
The most striking fact known at this date on the Green- Griffiths-Lang conjec-
ture is a recent result of Diverio, Merker and Rousseau [DMRIO] in the absolute
case, confirming the statement when X C ^^'^^ is a generic non singular hyper-
surface of large degree d, with a (non optimal) sufficient lower bound (i ^ 2" .
Their proof is based in an essential way on a strategy developed by Siu [Siu02,
Siu04], combined with techniques of [Dem95]. Notice that if the Green-Griffiths-
Lang conjecture holds true, a much stronger and probably optimal result would
be true, namely all smooth hypersurfaces of degree d ^ n + 3 would satisfy the
expected algebraic degeneracy statement. Moreover, by results of Clemens [Cle86]
and Voisin [Voi96], a (very) generic hypersurface of degree d ^ 2n + 1 would in
fact be hyperbolic for every n ^ 2. Such a generic hyperbolicity statement has
been obtained unconditionally by McQuillan [McQ98, McQ99] when n = 2 and
d ^ 35, and by Demailly-El Goul [DEGOO] when n = 2 and d ^ 21. Recently
Diverio- Trapani [DTIO] proved the same result when n = 3 and d ^ 593. By defi-
nition, proving the algebraic degeneracy means finding a non zero polynomial P
on X such that all entire curves f : C ^ X satisfy P{f) — 0. All known methods
of proof are based on establishing first the existence of certain algebraic differen-
tial equations P{f ; /', . . . , f^''^) = of some order k, and then trying to find
enough such equations so that they cut out a proper algebraic locus Y C X.
Let JkV be the space of fc-jets of curves / : (C, 0) — )■ X tangent to V. One
defines the sheaf 0{E^^V*) of jet differentials of order k and degree m to be
the sheaf of holomorphic functions P{z; ^i, . . .^k) on J^V which are homogeneous
polynomials of degree m on the fibers of J'^F — > X with respect to local coordinate
derivatives = f^^\0) ( see below in case V has singularities). The degree m
considered here is the weighted degree with respect to the natural C* action on J^'V
defined by A • f{t) := /(At), i.e. by rcparametrizing the curve with a homothetic
change of variable. Since (A • /)*•-''' (t) = f^^\Xt), the weighted action is given in
coordinates by
(0.3) A • (Ci, 6, • • • , 6) = (ACi, A'^a).
One of the major tool of the theory is the following result due to Green-Griffiths
[GG79] (see also [Blo26], [Dem95, Dem97], [SY96a, SY96b], [Siu97]).
(0.4) Fundamental vanishing theorem. Let {X, V) be a directed projective
variety and f : (C,Tc) — )> {X,V) an entire curve tangent to V. Then for every
global section P G H^{X,E^^V* ® 0{-A)) where A is an ample divisor of X ,
one hasP{f-f'J",...J^'^)) = Q.
It is expected that the global sections of H^{X^ E^^V* ®0{—A)) are precisely
those which ultimately define the algebraic locus Y C. X where the curve / should
lie. The problem is then reduced to the question of showing that there are many
non zero sections of iy°(X, E'^^y* ® 0(— ^)), and further, understanding what
4
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
is their joint base locus. The first part of this program is the main result of the
present paper.
(0.5) Theorem. Let {X, V) he a directed projective variety such that Ky is big
and let A be an am,ple divisor. Then for k ^ 1 and 5 G Q_|_ small enough,
5 ^ c{logk)/k, the number of sections h'^{Xj E^^V* (g> 0{—mdA)) has maximal
growth, i.e. is larger that CkmP^^'^~^ for some m ^ rrik, where c, Ck > 0,
n = dimX and r = rankF. In particular, entire curves f : (C,Tc) — )■ {X,V)
satisfy (many) algebraic differential equations.
The statement is very elementary to check when r = rank V — 1, and therefore
when n = dimX = 1. In higher dimensions n ^ 2, only very partial results
were known at this point, concerning merely the absolute case V = Tx- In
dimension 2, Theorem 0.5 is a consequence of the Riemann-Roch calculation of
Green-Griffiths [GG79], combined with a vanishing theorem due to Bogomolov
[Bog79] - the latter actually only applies to the top cohomology group i?"^, and
things become much more delicate when extimates of intermediate cohomology
groups are needed. In higher dimensions, Diverio [Div09] proved the existence
of sections of H^{X,E^^V* ® 0(-l)) whenever X is a hypersiirfacc of PJ^
of high degree d ^ d^, assuming k ^ n and m ^ m^. More recently, Merker
[MerlO] was able to treat the case of arbitrary hypersurfaces of general type, i.e.
d ^ n + 3, assuming this time k to be very large. The latter result is obtained
through explicit algebraic calculations of the spaces of sections, and the proof is
computationally very intensive. Berczi [BcrlO] also obtained related results with
a different approach based on residue formulas, assuming d ^ 2^"^'°^'^.
All these approaches are algebraic in nature, and while they use some form
of holomorphic Morse inequalities [Dem85], they only require a very special
elementary algebraic case, namely the lower bound
for L = 0{A — B) with A, B ncf (cf. Trapani [Tra95]). Here, our techniques
are based on more elaborate curvature estimates in the spirit of Cowen- Griffiths
[CG76] . They require the stronger analytic form of holomorphic Morse inequalities.
(0.6) Holomorphic Morse inequalities ([Dem85]). Let (L, h) be a holomorphic
line bundle on a compact complex manifold X , equipped with a smooth hermitian
metric h, and let E be a holomorphic vector bundle. Denote by L,h = ~ irf ^
the curvature form of (L, h) and consider the open set
X{L, h,q) — [x & X ; QL,h{x) has signature {n — q, q)] {q-index set of QL,h)^
so that there is a partition X = S'U1Jq^^^^X(L, h, q) where S = {det ©L,ft(a;)=0}.
Then, if we put r = rankE', we have asymptotically as m tends to infinity:
(a) ( Weak Morse inequalities)
m
n
n!
0. Introduction 5
(b) {Strong Morse inequalities) If X{L,h,^ q) = Yij^q-^i^j^jj)) then
(c) {Lower hound on hP)
h^{X,E®L®'^)-h'-{X,E®L®'^)^'^r I Ql - o{m'') .
The proof of the above is based on refined spectral estimates for the complex
Laplace-Beltrami operators. Observe that (0.6 c) is just the special case of (0.6 b)
when q = 1. It has been recently observed that these inequalities should be
optimal in the sense that the asymptotic cohomology functional h'^{X,L) :=
limsup^^+^ ^/iO(X,L®-) satisfies
(0.7) hi{x,L)^ inf [ {-lye
n
X{L,h,q)
and that conjecturally the inequality should be an equality; it is proved in
[DemlOa], [DemlOb] that this is indeed the case if n ^ 2 or g = 0, at least
when X is projective algebraic.
Notice that holomorphic Morse inequalities are essentially insensitive to sin-
gularities, as we can pass to non singular models and blow-up X as much as we
want: if : X — > X is a modification then /i*0~ = Ox and R'^ji^^^Q- is sup-
ported on a codimension 1 analytic subset (even codimension 2 if X is smooth).
It follows by the Leray spectral sequence that the estimates for L on X or for
L — iJ,*L on X differ by negligible terms 0{m^~^). Finally, we can even work
with singular hermitian metrics h which have analytic singularities with positive
rational coefficients, that is, one can write locally h = e"*^ where, possibly after
blowing up,
(0.8) Lp{z) = clog^^ mod C°°, with c e Q_|_ and gj holomorphic.
j
Especially, (p is smooth on some Zariski open set X \ Z where Z = f]g~\0),
and it has logarithmic poles along Z. Blowing-up the ideal sheaf d = {gj) leads
to divisorial^ singularities, and then by replacing L with L = ii*L ® Q{—E) where
E G DivQ(X) is the singularity divisor, we see that holomorphic Morse inequalities
still hold for the sequence of groups Hi{X, E(g)L®'^(g)3{h®'^)) where J(/i®'^) is the
multiplier ideal sheaf of h®"^ (see Bonavero [Bon93] for more details). In the case
of linear subspaces V C Tx, we introduce singular hermitian metrics as follows.
(0.9) Definition. A singular hermitian metric on a linear subspace V C Tx is
a metric h on the fibers of V such that the function log/i : ^ h- )■ log |^|^ is locally
integrable on the total space ofV.
6
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
Such a metric can also be viewed as a singular hermitian metric on the tauto-
logical line bundle Op(x/)(— 1) on the projectivized bundle P{V) — V \ {0}/C*,
and therefore its dual metric h* defines a curvature current ©Op(v)(i),/i* type
(1, 1) on P{V) C P(Tx), such that
P*©o,(.)(i),h* = ^ddlogh, where p : F \ {0} ^ P{V).
If log h is quasi-plurisubharmonic (or quasi-psh, which means psh modulo addition
of a smooth function) on V, then \ogh is indeed locally integrable, and we have
moreover
(0.10) ©Op(,)(i),ft* > -Cu
for some smooth positive (1, l)-form on P{V) and some constant C > ;
conversely, if (0.10) holds, then log/i is quasi psh.
(0.11) Definition. We will say that a singular hermitian metric h on V is
admissible if h can be written as h = e'^hQ^y where ho is a smooth positive definite
hermitian on Tx and (p is a quasi-psh weight with analytic singularities on X , as
in (0.9). Then h can he seen as a singular hermitian metric on Op(v')(l), with
the property that it induces a smooth positive definite metric on a Zariski open set
X' C X \ Sing(y) ; we will denote by Sing(/i) D Sing(y) the complement of the
largest such Zariski open set X' .
If h is an admissible metric, we define OhiV*) to be the sheaf of germs of
holomorphic sections sections of ^x\Sing(/i) which are /i*-bounded near Sing(/i);
by the assumption on the analytic singularities, this is a coherent sheaf (as the
direct image of some coherent sheaf on P{V)), and actually, since h* — e~'^hQ,
it is a subsheaf of the sheaf 0{V*) :— Of^XV*) associated with a smooth positive
definite metric ho on Tx- If r is the generic rank of V and m a positive
integer, we define similarly Kyf^ to be sheaf of germs of holomorphic sections
of (dct Vf^,)'^"' = (A'^1^^,)®"' which are det /i*-bounded, and Kl^ := K^^^^.
If V is defined hy a : X —-> Gr{Tx), there always exists a modification
H : X ^ X such that the composition ao^ : X — )■ Gr{n*Tx) becomes holomorphic,
and then /x* V|^-i(X\Sing(\^)) extends as a locally trivial subbundle of /x*Tx which
we will simply denote by fj,*V. If h is an admissible metric on V, then fj,*V can
be equipped with the metric iJ,*h = e'^°^iJ,*ho where //*/io is smooth and positive
definite. We may assume that tp o ^ has divisorial singularities (otherwise just
perform further blow-ups of X to achieve this). We then see that there is an integer
mo such that for all multiples m = pruQ the pull-back iJb*Kyy^ is an invertible sheaf
on X, and deth* induces a smooth non singular metric on it (when h = Hq, we
can even take mo = 1). By definition we always have Ky^ = fi^{fi*KYiJ for any
m ^ 0. In the sequel, however, we think of Kv,h not really as a coherent sheaf,
but rather as the "virtual" Q-line bundle //*(//*K^° )i/"^o, and we say that Kv,h
is big if h^{X,Ky^ ^ cm"' for m ^ mi, with c > 0, i.e. if the invertible sheaf
ji*Ky°y^ is big in the usual sense.
0. Introduction 7
At this point, it is important to observe that "our" canonical sheaf Ky differs
from the sheaf %v '■= i^,0{Kv) associated with the injection i : X \Smg{V) )■ X,
which is usuaUy referred to as being the "canonical sheaf" , at least when V is the
space of tangents to a foliation. In fact, %v is always an invertible sheaf and there
is an obvious inclusion Ky C Xy- More precisely, the image of 0(A''TJ) — > Xy is
equal to Xy ®0x 3 for a certain coherent ideal 3 C Ox, and the condition to have
/iQ-bounded sections on X\Sing(y) precisely means that our sections are bounded
by Const^ l^f^ l in terms of the generators (gj) of Xy 3, i-e. Ky — Xy 3
where 3 is the integral closure of d- More generally,
where C Ox is the (m/mo)-integral closure of a certain ideal sheaf
dh,mo C Ox, which can itself be assumed to be integrally closed; in our previous
discussion, // is chosen so that iJ,*3h,mo is invertible on X.
The discrepancy already occurs e.g. with the rank 1 linear space V C Tp»i
consisting at each point z ^ of the tangent to the line (0^) (so that necessarily
Vq — Tp^^o)- As a sheaf (and not as a linear space), i*0(F) is the invertible sheaf
generated by the vector field ^ = '^Zjd/dzj on the affine open set C P^, and
therefore Xy :— i^O{V*) is generated over C"^ by the unique 1-form u such that
u{^) = 1. Since ^ vanishes at 0, the generator u is unbounded with respect to
a smooth metric hg on Tp», and it is easily seen that Ky is the non invertible
sheaf^Ky = Xy ® mp",o- We can make it invertible by considering the blow-up
: X ^ X oi X = at 0, so that fx*Ky is isomorphic to fi*Xy (g) 0~{-E)
where E is the exceptional divisor. The integral curves C of are of course
lines through 0, and when a standard parametrization is used, their derivatives do
not vanish at 0, while the sections of i*0{V) do - another sign that i*0{V) and
i*0(y*) are the wrong objects to consider. Another standard example is obtained
by taking a generic pencil of elliptic curves XP{z) + iJiQ{z) = of degree 3 in P^,
and the linear space V consisting of the tangents to the fibers of the rational map
Pc defined by ^ H> Q{z)/P{z). Then V is given by
i,0{V) 0(Tp2) ''^^~^'^'') Op2(6) as ^
where S = Smg{V) consists of the 9 points {P{z) = 0}n{Q{z) — 0}, and ds is the
corresponding ideal sheaf of S. Since det 0(Tp2) = 0(3), we see that Xy = 0(3) is
ample, which seems to contradict (0.2) since all leaves are elliptic curves. There is
however no such contradiction, because Ky = Xy ® 3s is not big in our sense (it
has degree on all members of the elliptic pencil). A similar example is obtained
with a generic pencil of conies, in which case Xy = 0(1) and cardS* = 4.
For a given admissible hermitian structure {V,h), we define similarly the
sheaf E^^V}^ to be the sheaf of polynomials defined over X \ Sing(/i) which
are "/i-bounded" . This means that when they are viewed as polynomials
8
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
P{z; ^1, . . . ,^k) in terms of = {^h^YfiO) where V)j'° is the (1, 0)-component
of the induced Chern connection on (y,ho), there is a uniform bound
(0.12) \P{z;^i,...,^k)\^c[Y.Uj\\l/'y
near points of X \ X' (see section 2 for more details on this). Again, by a direct
image argument, one sees that E^^Vj^ is always a coherent sheaf. The sheaf
E^^V* is defined to be E^'^Vj^ when h = Kq (it is actually independent of the
choice of /iq, as follows from arguments similar to those given in section 2). Notice
that this is exactly what is needed to extend the proof of the vanishing theorem 0.4
to the case of a singular linear space V ; the value distribution theory argument
can only work when the functions P(/ ; /',..., f^^^){t) do not exhibit poles, and
this is guaranteed here by the boundedness assumption.
Our strategy can be described as follows. We consider the Green- Griffiths
bundle of /c-jets X^'^ = J^V \ {0}/C*, which by (0.3) consists of a fibration in
weighted projective spaces, and its associated tautological sheaf
L = 0^gg(1),
A:
viewed rather as a virtual Q-line bundle Oj^^gg (mo)^/"^" with mo = lcm(l, 2, ... , k).
Then, if tt^ : — )■ X is the natural projection, we have
^k'm ~ {'^k)*OxGG{m) and i?'^(7rfc)*0;sfGG (m) = for g ^ 1.
' k k
Hence, by the Leray spectral sequence we get for every invertible sheaf F on X
the isomorphism
(0.13) H'^{X,E^^^V*®F) ^ Jf^^(X,^'«,O^GG(m) ® TT^F).
' k
The latter group can be evaluated thanks to holomorphic Morse inequalities. In
fact we can associate with any admissible metric h on V a, metric (or rather a
natural family) of metrics on L = 0;!^gg(1). The space X^^ always possesses
quotient singularities if /c ^ 2 (and even some more if V is singular), but we
do not really care since Morse inequalities still work in this setting. As we will
see, it is then possible to get nice asymptotic formulas as k ^ +oo. They
appear to be of a probabilistic nature if we take the components of the fc-jet
(i.e. the successive derivatives = /'■"''' (0), 1 ^ j ^ k) as random variables.
This probabilistic behaviour was somehow already visible in the Riemann-Roch
calculation of [GG79] . In this way, assuming Ky big, we produce a lot of sections
aj = H'^(X^'^ ,OxGG{m) (8) tt^F), corresponding to certain divisors Zj C X^^.
The hard problem which is left in order to complete a proof of the generalized
Green-Griffiths-Lang conjecture is to compute the base locus Z = f]Zj and to
show that Y = nk{Z) C X must be a proper algebraic variety. Although we
cannot address this problem at present, we will indicate a few technical results
and a couple of potential strategies in this direction.
I would like to thank Simone Diverio and Mihai Paun for several stimulating
discussions, and Erwan Rousseau for convincing me to explain better the pecu-
liarities of the definition of the canonical sheaf employed here.
1. Hermitian geometry of weighted projective spaces 9
1. Hermitian geometry of weighted projective spaces
The goal of this section is to introduce natural Kahler metrics on weighted
projective spaces, and to evaluate the corresponding volume forms. Here we put
d'^ = -^{d — d) so that dd^ = ^dd. The normalization of the d^ operator is chosen
such that we have precisely {dd'^ log \ z\'^)"' — Sq for the Monge- Ampere operator
in C"; also, for every holomorphic or meromorphic section cr of a hermitian line
bundle (L, h) the Lelong-Poincare can be formulated
(1.1) dd''log\a\l^[Z^]-eL,h,
where 0L,h = 2^-D| /j is the (1, l)-curvature form of L and the zero divisor of cr.
The closed (1, l)-form OL,h is a representative of the first Chern class ci(L). Given
a fc-tuple of "weights" a = (ai, . . . , a^), i.e. of relatively prime integers > we
introduce the weighted projective space P{ai, . . . , a^) to be the quotient of C'^xjO}
by the corresponding weighted C* action:
(1.2) P(ai, . . . , afe) = {0}/C*, X-z = (A^^^i, . . . , X'^'^Zk).
As is well known, this defines a toric k — 1-dimensional algebraic variety with
quotient singularities. On this variety, we introduce the possibly singular (but
almost everywhere smooth and non degenerate) Kahler form u!a,p defined by
(1.3) 7r*UJa,p = dd''(fa,p, <fa,pi^)=-^Og J2 1^*1^^^'*%
where tTq : C'^ \ {0} — )> P(ai, . . . , Ofe) is the canonical projection and p > is
a positive constant. It is clear that (pp^a is real analytic on C'^ \ {0} if p is an
integer and a common multiple of all weights a^. It is at least is p is real and
p ^ max(as), which will be more than sufficient for our purposes (but everything
would still work for any p > 0). The resulting metric is in any case smooth
and positive definite outside of the coordinate hyperplanes = 0, and these
hyperplanes will not matter here since they are of capacity zero with respect to
all currents {dd'^(pa,pY ■ In order to evaluate the volume /pj-^j^ ak)^o,^^
observe that
J P{ai,...,ak) ^2eC'=,(^a,p(2)=0
= / {dd^Va,pf~^ ^d^^a.p
'2eC'=,V5a,p(2)=0
(1.4) = 4 / {dd'^e^'^'^'^f.
The first equality comes from the fact that {'Pa,p{z) = 0} is a circle bundle
over P(ai, . . . , afc), by using the identities (fa,pi^ " = Va,p{z) + log|Ap and
10 Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
i|A|=i d^^^S I'^P = 1- The third equahty can be seen by Stokes formula applied to
the {2k - l)-form
{dcTe^'^'^'^f-'^ A d^'e'P'f-'^ = e^'f-'^{dd^^a,pf~^ A d^a.p
on the pseudoconvex open set e C'^ ; (pa,p{z) < 0}. Now, we find
(1.5) {dd'e^'^'^'^f ^ {dd" \zs\^P'''^^^ ^ JJ (^l.\z^\-^.-^Ydd''\z\''f,
(1.6) / {dd^e^^^'^f = TT ^ = .
AeCfc,<^a,p(2)<o i^skk^' ai...ak
In fact, (1.5) and (1.6) are clear when p — ai = . . . = ak = 1 (this is just the
standard calculation of the volume of the unit ball in C'^); the general case follows
by substituting formally Zg z^J"'% and using rotational invariance along with
the observation that the arguments of the complex numbers t^J""" now run in the
interval [0,27rp/as[ instead of [0, 27r[ (say). As a consequence of (1.4) and (1.6),
we obtain the well known value
(1.7) / a;,^-^- ^
p(ai,...,ofc) ai...afe
for the volume. Notice that this is independent of p (as it is obvious by Stokes
theorem, since the cohomology class of oj^^p does not depend on p). When p tends
to +00, we have ^Pa,p{z) i->- <^a,oo(^) = logmaxi^g^/- Izsp/'*'' and the volume form
^\~p converges to a rotationally invariant measure supported by the image of the
polycircle Illksl — 1} -P(tti, . . . , a/c). This is so because not all l^sp/"^ are
equal outside of the image of the polycircle, thus (/9a,oc(-2) locally depends only
on A; — 1 complex variables, and so o;^"^ = there by log homogeneity.
Our later calculations will require a slightly more general setting. Instead of
looking at C'^, we consider the weighted C* action defined by
(1.8) Cl'^l = X . . . X C'^^ A-^ = (A"i2i,...,A"*2fc).
Here Zg G for some /c-tuple r = (ri, . . . , r^) and |r| = ri + . . . + r/e. This gives
rise to a weighted projective space
P(af . . . , a}r'=^) = P(ai, . . . , ai, . . . , a^, . • . , afc),
(1.9) 7ra,r-:C^^ X ... xC^'^ \{0} ^P(a^],...,aJ:'=])
obtained by repeating times each weight a^. On this space, we introduce the
degenerate Kahler metric ijJa,r,p such that
(1.10) 7rl^U;a,r,p = dd^(pa,r,p, (pa,r,p{^) ^ - log ^ l^sP^^''^
1. Hermitian geometry of weighted projective spaces
11
where l^:^! stands now for the standard hermitian norm (^i<^j^^^
This metric is cohomologous to the corresponding "polydisc-like" metric 0Ja,p
already defined, and therefore Stokes theorem implies
(1-11) / , , , ,
r|-l
r 1 r 1 OiJ^iP Tl fh
Since {dd'^ log |2;sp)'^'' = on C^" \{0} by homogeneity, we conclude as before that
the weak limit limp_^+oo ^^a}r,p = '^a,r,oo associated with
(1.12) (Pa,r,oo{z) = log max \z.
s
|2/a.
is a measure supported by the image of the product of unit spheres J| S'^'^''"-'^
in P{a^p\ . . . , a['^''^), which is invariant under the action of U{ri) x . . . x U(rk)
on X ... X C^* , and thus coincides with the hermitian area measure up to a
constant determined by condition (1.11). In fact, outside of the product of spheres,
fa.r,oo locally depends only on at most k — 1 factors and thus, for dimension
reasons, the top power {dd'^(fia,r,oo)^^^~^ must be zero there. In the next section, the
following change of variable formula will be needed. For simplicity of exposition
we restrict ourselves to continuous functions, but a standard density argument
would easily extend the formula to all functions that are Lebesgue integrable with
respect to the volume form uja}r,p-
(1.13) Proposition. Let f{z) be a bounded function on P(a^],...,a^l ) which
is continuous outside of the hyperplane sections Zg = 0. We also view f as a
C* -invariant continuous function on J|(C^^ \ {0}). Then
r -1
T->/ ['^l ] [^h\\
r.,-1
= M_i^ / fixr/'%„...,xi^/'%,) n j^^^.dxd^^iu)
where A/-_i is the {k — l)-simplex {xg ^ 0, ^ = 1}, dx — dxi A ... A dx^-i its
standard measure, and where djjiiu) — dniiui) . . . dnk{uk) is the rotation invariant
probability measure on the product Ylg S^^''~^ of unit spheres in C^^ x . . . x C'"'=.
As a consequence
lim / f{z)'^al~r!' = T-r^ r I f{u)dlJ,(u).
Proof. The area formula of the disc J..^^^^dd^\X\'^ = 1 and a consideration of the
unit disc bundle over P{a^p\...,a^^'''^) imply that
Ip--= I , , , ,/(^)<l7/= / f{z){dd'^ipa,rJ'^^-'^dd'^e^^^^^^.
12 Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
Now, a straightforward calculation on Cl^' gives
On the other hand, we have {dd''\z\'^y'^\ — Yli^g^ki^^^l^sl'^y" and
thanks to the homogeneity relation {dd'^ipa,r,p)^^^ = 0. Putting everything together,
we find
A standard calculation in polar coordinates with Zs = PsUg, Ug G S'^^^"-'^, yields
{dd''\zs\'^y^ dps , .
where fis is the t/(rs)-invariant probability measure on S'^''""^. Therefore
J ^ f (|r| - l)!p^-V(pi^i, ■ ■ ■ , p,u,) yr 2pr-/"- ^d^^sM
^ f (|r| - i)!j9-v(tr^'^^i, • • • , c^'^'^fc) n ^^^"'^^^
by putting = l^spP/"" = pl^^""", i.e. ps = t"^^^^, G ]0, 1]. We use still another
change of variable ts = tXs with t = Yji^s^k^s and Xs G ]0, 1], Ei^s^fc^s = 1-
Then
dti A ... A dtfe = t'^""'^ dx dt where dx = dx\ A ... A dxk-i-
The C* invariance of / shows that
x'^/-^dp,s{us)
/„ =
dx.
2. Probabilistic estimate of the curvature of A;-jet bundles 13
This is equivalent to the formula given in Proposition 1.13. We have xt^^^ — )■ 1 as
p — )■ +00, and by Lebesgue's bounded convergence theorem and Fubini's formula,
we get
lim /„ = -^J^l — ^ I f(u) TT — —rdxdaiu).
It can be checked by elementary integrations by parts and induction on A;, ri, . . . ,rk
that
(1.14) / n <'~'dx^...dx,.,^jj-^^ n (^^-1)'-
This implies that (|r| — 1)! Hi^s^fe (^"-1)! ^ probability measure on Afe_i and
that
1
Even without an explicit check, the evaluation (1.14) also follows from the fact
that we must have equality for f{z) = 1 in the latter equality, if we take into
account the volume formula (l-ll). □
2. Probabilistic estimate of the curvature of fc-jet bundles
Let {X, V) be a compact complex directed non singular variety. To avoid any
technical difficulty at this point, we first assume that V is a holomorphic vector
subbundle of Tx, equipped with a smooth hermitian metric h.
According to the notation already specified in the introduction, we denote by
J'^V the bundle of /c-jets of holomorphic curves / : (C, 0) — )■ X tangent to V at
each point. Let us set n = dimcX and r = ranked. Then J^V ^ X is an
algebraic fiber bundle with typical fiber C^'^ (see below). It has a canonical C*-
action defined by A • / : (C, 0) — )■ X, (A • f){t) — /(At). Fix a point xq in X and
a local holomorphic coordinate system (^i, . . . , Zn) centered at xq such that V^,, is
the vector subspace {d/dzi, . . . , d/dzr) at xq. Then, in a neighborhood U of xq,
V admits a holomorphic frame of the form
(2.1) — + V a^^piz)--, l^p^r, a^p{0) = 0.
OZp .f-^^ OZa
Let f{t) — {fi{t),...,fn{t)) be a k-jet of curve tangent to V starting from a
point /(O) = X E U. Such a curve is entirely determined by its initial point
and by the projection f{t) := {fi{t), . . . , fr{t)) to the first r-components, since
the condition f'{t) G ^fit) implies that the other components must satisfy the
ordinary diff'erential equation
14
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
This implies that the fc-jet of / is entirely determined by the initial point x and
the Taylor expansion
(2.2) fit) -x = ^^t + + . . . + + C'(i'+')
where = (^sa)i^a^r £ C*^. The C* action (A,/) i-^ A • / is then expressed in
coordinates by the weighted action
(2.3) A • (Ci, 6, . . . , efc) = (Aei, A^6, • • • , A^Cfc)
associated with the weight a = 2^'^\ . . . , /c^). The quotient projectived k-jet
bundle
(2.4) X^^ := (J'^V \ {0})/C*
considered by Green and Griffiths [GG79] is therefore in a natural way a
p(^lM^ 2M, . . . , fcM) weighted projective bundle over X. As such, it possesses a
canonical sheaf Oj^GG(f) such that OxGG(m) is invertible when m is a multiple of
lcm(l, 2, . . . , /c). Under the natural projection tt^ : X^'^ — )■ X, the direct image
(7rfc)*0^GG (m) coincides with the sheaf of sections of the bundle E^'^V* of jet
differentials of order k and degree m, namely polynomials
(2.5) P{z;Ci,...,Ck)^ E «ai...a.(^)^r---C-
of weighted degree |q;i| + 2\a2\ + ... + k\ak\ = m on J^V with holomorphic
coefficients. The jet differentials operate on germs of curves as differential
operators
(2.6) Pirn) = f'itr ■ --f^'Htr
In the sequel, we do not make any further use of coordinate frames as (2.1), because
they need not be related in any way to the hermitian metric h of V. Instead, we
choose a local holomorphic coordinate frame {ect{z))i^a^r of F on a neighborhood
U of xo, such that
(2.7) {ea{z),ei3{z)) = Sal3 + Cijal3ZiZj + 0{\z\^)
for suitable complex coefficients (cija/s). It is a standard fact that such a
normalized coordinate system always exists, and that the Chern curvature tensor
^Dyh (^' ^) -^0 is then given by
(2.8) Ov,h{^o) = X] (^ija(3 dzi A dzj ®el^®ep.
2. Probabilistic estimate of the curvature of A;-jet bundles
15
Also, instead of defining the vectors e C as in (2.2), we consider a local
holomorphic connection V on V\u (e.g. the one which turns (cq.) into a parallel
frame), and take = ^^/(O) G defined inductively by V^/ = /' and
V^/ = V/'(V^~^/). This is just another way of parametrizing the fibers of J'^V
over U by the vector bundle Notice that this is highly dependent on V
(the bundle J''V actually does not carry a vector bundle or even affine bundle
structure); however, the expression of the weighted action (2.3) is unchanged
in this new setting. Now, we fix a finite open covering {Ua)aei of ^ by open
coordinated charts such that V\u^ is trivial, along with holomorphic connections
Vq, on Let be a partition of unity of X subordinate to the covering (Ua)-
Let us fix p > and small parameters 1 = £i S> £2 ^ • • • ^ ^fe > 0. Then we
define a global weighted exhaustion on J^V by putting for any /c-jet f e J^V
(2.9) *h,p,s(/):=(E^"(^) E ^ini'^'jmlt^f'
where || \\h{x) is the hermitian metric h of V evaluated on the fiber V^, x = /(O).
The function '^h,p,£ satisfies the fundamental homogeneity property
(2.10) ^h,pA^-f) = ^h,pAf)\M^
with respect to the C* action on J'^V^ in other words, it induces a hermitian metric
on the dual L* of the tautological Q-line bundle = Oxgg{1) over X^*^. The
curvature of is given by
(2-11) nieL„^;^^^=dd'log<i^H,p,e
where ttj. J^V \ {0} — )■ X^*^ is the canonical projection. Our next goal is
to compute precisely the curvature and to apply holomorphic Morse inequalities
to L ^ X^^ with the above metric. It might look a priori like an untractable
problem, since the definition of '^h,p,e is a rather unnatural one. However, the
"miracle" is that the asymptotic behavior of '^h,p,e as Sg/sg-i ^ is in some
sense uniquely defined and very natural. It will lead to a computable asymptotic
formula, which is moreover simple enough to produce useful results.
(2.12) Lemma. On each coordinate chart U equipped with a holomorphic connec-
tion V of V\u, let us define the components of a k-jet f G J^V by = V®/(0),
and consider the resettling transformation
P£(Ci,6,---,Cfe) = (eki,ei6,---,4Cfe) on J^V, xeUa
{it commAites with the C* -action but is otherwise unrelated and not canonically
defined over X as it depends on the choice of V). Then, if p is a multiple
of \cm{l, 2, k) and Ss/Ss-i — >■ for all s = 2,...,k, the res caled function
^h,p,e o PrHCi> • • • , Cfe) converges towards
on every compact subset of J^V^u \ {0}, uniformly in C°° topology.
16
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
Proof. Let U C X be an open set on which Vj^/ is trivial and equipped with
some holomorphic connection V. Let us pick another holomorphic connection
V = V + r where F e H^{U, Q]^ Hom(F, V). Then V^/ = V^/ + r(/)(/') ■
and inductively we get
vv = vv + n(/; vV,...,v^-V)
where P{x ; ^i, . . . ,^s-i) is a polynomial with holomorphic coefficients in x & U
which is of weighted homogeneous degree s in (^i, . . . , ^<,_i). In other words,
the corresponding change in the parametrization of J^V^u is given by a C*-
homogeneous transformation
l8 = ia + Pa{x] ii,...,^s-l)-
Let us introduce the corresponding rescaled components
(Cl,e>--->Cfc,e) = {£lil,---,£k^k), (Cl,e>---,Cfc,e) = • • • » 4Cfc)-
Then
= ^s,e + OiSs/Ss-lY 0(||6,e|| + • • • + f Z^^"'^)^
and the error terms are thus polynomials of fixed degree with arbitrarily small
coefficients as Ss/ss-i — > 0. Now, the definition of '^h,p,s consists of glueing the
sums
corresponding to = V^/(0) by means of the partition of unity ^ Ooi{x) — 1. We
see that by using the rescaled variables ^s,e the changes occurring when replacing a
connection Vq by an alternative one V/3 are arbitrary small in C°° topology, with
error terms uniformly controlled in terms of the ratios Ss/ss-i on all compact
subsets of \ {0}. This shows that in C°° topology, ^h,p,£ ° Pj^i^iT ■ ■ ^^k)
converges uniformly towards (Xli<s<fc ll^fcll^^^*)^^^' whatever is the trivializing
open set U and the holomorphic connection V used to evaluate the components
and perform the rescaling. □
Now, we fix a point Xq E X and a local holomorphic frame {eot{z))i^a^r
satisfying (2.7) on a neighborhood U of xq. We introduce the rescaled components
= esV*/(0) on J^V\u and compute the curvature of
*/.,p,sopr'(^;ei,...,a)^ ( E ii^^ii'^'^')
1/p
(by Lemma 2.12, the errors can be taken arbitrary small in C°° topology). We
write = Yli^a^r^saea- By (2.7) we have
lle.|l^ = El^««l'+ E CijapZiZjU'^s^ + 0{\z\^\C\^).
2. Probabilistic estimate of the curvature of A;-jet bundles
17
The question is to evaluate the curvature of the weighted metric defined by
*(^;ei,...,6) = ( E iie^iiH
= ( E (Ei^^«i' + E ^^^"/j^^^i^^^M )
1/P
We set l^sp = X^Q, l^sap. A straightforward calculation yields
log^'(2;; ^1,...,^) =
By (2.11), the curvature form of — Oxoa (1) is given at the central point xq by
the following formula.
(2.13) Proposition. With the above choice of coordinates and in terms of the
resettled components — V^/(0) at xq E X , we have the approximate expression
^L„^l,,J''0,m-^a,r,p{0+^ E 7 v 1^ \2p/t E C^jaP^T^ ^
where the error terms are 0{m.ax.2<^ss^k{^s/ ^s-iY) uniformly on the compact
variety X^*^ . Here u)a,r,p is the (degenerate) Kdhler metric associated with the
weight a = (1^, 2^, . . . , fcM) of the canonical C* action on J^V .
Thanks to the uniform approximation, we can (and will) neglect the error
terms in the calculations below. Since uja,r,p is positive definite on the fibers of
X^^ X (at least outside of the axes — 0), the index of the (1, 1) curvature
form QLk,^l {^i [C]) is equal to the index of the (1, l)-form
(2.14) ^^(z,i):^^ E \^HfWrt E c,Jocp{^)^^^dzi^dz_
depending only on the differentials (<^-2j)i^j^n on X. The qf-index integral of
(Lfe, p g) on X'^'^ is therefore equal to
/ r\n-\-kr—l
(n + fcr-l)! [ f <-;(0l7.,.(^,07fe(^,er
n\{kr - 1)! J^^x J^eP{iM,-,kM)
18
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
where lyi^^q{z,^) is the characteristic function of the open set of points where
7fc(2;,^) has signature (n — q,q) in terms of the dzj^s. Notice that since 7^(2,^)"^
is a determinant, the product l^i^^q{z,^)'^k{z,0"' gives rise to a continuous function
on X^*^. Formula 1.13 with n = . . . = rfc = r and = s yields the slightly more
explicit integral
\r-l
+kr-i ^ (n + fcr- 1)!
f f hk,qi^i X, u)gk{z, X, uY ' ' ' j — dxdii(u),
izex J(x,«)eAfc_ix(S2r- i)fe (r-lj!
where 5ffe(2;, x, u) = jk{z, x\^'^^ui, x^/'^^Uk) is given by
i . 1
(2.15) gk{z,x,u) ^ — ^ -Xg ^ Cijc,i3{z)usaUspdzi Adzj
and ]lg^.,q(2, tt) is the characteristic function of its g-index set. Here
(2.16) dukAx) = {kr - 1)! ^''(^•:''^')|r'
is a probability measure on Aj._i, and we can rewrite
f on+fer-l ^ (n + fcr- 1)!
yx-(L„,) ""^'^^^ n!(fc!)'-(fcr-l)!
(2.17) / / lg^Az,x,u)gk{z,x,uy duk,r{x)dii{u).
JzeX 7(a;,M)eAfe_ix(S2'-i)fe
Now, formula (2.15) shows that gk{z,x,u) is a "Monte Carlo" evaluation of
the curvature tensor, obtained by averaging the curvature at random points
Us G 5'^''"^ with certain positive weights Xg/s; we should then think of the
fc-jet / as some sort of random parameter such that the derivatives V'^/(0) are
uniformly distributed in all directions. Let us compute the expected value of
{x,u) gk{z^ X, u) with respect to the probability measure duk^rix) dfj,{u). Since
Js^r-i UsaUs/3dfi{us) = ^6^/3 and /^^_^ Xg dv^^x) = \, we find
E(5rfc(^,», •)) = ^ ■ -^^Cijaa{z)dZi Ad:zj.
In other words, we get the normalized trace of the curvature, i.e.
(2.18) E{gkiz,.,.)) = + ^ + ... + ^)Qdet{V*),deth*,
2. Probabilistic estimate of the curvature of A;-jet bundles
19
where Odet(\/*),det h* is the (1, l)-curvature form of det(y*) with the metric induced
by h. It is natural to guess that gk{z, x, u) behaves asymptoticaUy as its expected
value Ei{gk{z, •)) when k tends to infinity. If we replace brutally gk by its
expected value in (2.17), we get the integral
(n + A;r — 1)!
n\{k\y{kr
)! 1 / 1 r
where i] := ©det(v*),det/i* ^ri,q is the characteristic function of its g-index
set in X. The leading constant is equivalent to (log /c)"'/n!(A;!)^ modulo a
multiplicative factor 1 + 0(1/ log A;). By working out a more precise analysis of
the deviation, we will prove the following result.
(2.19) Probabilistic estimate. Fix smooth hermitian metrics h on V and
— 27r Tli^ijd'^i ^ d^j on X . Denote by Qv,h = S c-ijapdzi A dzj ® e*^®
the curvature tensor ofV with respect to an h-orthonormal frame (cq), and put
77(2;) = 0det(V^*),deth* = 1^ XI VijdZiAdZj, r]ij = X Cijaa-
Finally consider the k-jet line bundle — Oxgg;(1) equipped with the
induced metric p e (^^ defined above, with 1 = £1 ^ £2 ^ • • • fffc > 0).
When k tends to infinity, the integral of the top power of the curvature of on
its q-index set X^^(Lfe, q) is given by
L
for all q = 0,1, ... ,n, and the error term 0((log k) ^) can be bounded explicitly in
terms of Qy , V O'lT'd oj. Moreover, the left hand side is identically zero for q > n.
The final statement follows from the observation that the curvature of is
positive along the fibers of X^'^ X, hj the plurisubharmonicity of the weight
(this is true even when the partition of unity terms are taken into account, since
they depend only on the base); therefore the index sets are empty for q > n. We
start with three elementary lemmas.
(2.20) Lemma. The integral
Ik,r,n = / [ ^ ~] dVk,r{x)
is given by the expansion
(a) Ik,r,n — X
(kr-iy. Ui^i^kir - ^ + PiV-
Ks. so 3 <fcSiS2...Sn (r-l)!'= {kr + n-iy.
20 Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
where j3i = Pi{s) = card{j ; Sj = i}, Pi = n, 1 ^ i ^ k. The quotient
I / 1
kr{kr + l)...{kr + n-l)\ ^2^"'^ k)
is bounded below by 1 and bounded above by
m=2 V ^ \ /
As a consequence
(C) h,r,n =^{{} + \ + ---+\Y + Oiilogkr-'))
_ (log + 7)'" + 0((log k)""-^)
~ k^
where 7 is the Euler-Mascheroni constant.
Proof. Let us expand the n-th power (X^i^g^fc T")"" '^^^ gives
Ik,r,n — ^ ^ ~s~S S~ ^ "^^^ ' ' dvkjri^)
and by definition of tlie measure Uk^r we liave
/ . . . = ^^^1^ / xl+^^-' . . . 4+^^=-^ dx,... dXk.
By Formula (1-14), we find
/ x^^ x^^d.. (X) nK.^.(^+A-i)'
_ ^^n,;3.^i(i + ^)(i + f)---(i + V)
kr{kr + 1) . . . {kr + n — 1) '
and (2.20 a) follows from the first equality. The final product is minimal
when r = 1, thus
kr{kr + l)...{kr + n-l)^ j ' ' ' '^'''^'"^^^
(2.21) ^
kr{kr + 1) . . . {kr + n — 1)
2. Probabilistic estimate of the curvature of A;-jet bundles
21
Also, the integral is maximal when all Pi vanish except one, in which case one gets
f ^ , .X r(r + 1) . . . (r + n — 1)
(2.22) / x]duk,r{x)- ^ ' ^ '
kr{kr + 1) . . . (kr + n — 1)
By (2.21), we find the lower and upper bounds
(2.23) ^'''-'-^ kr{kr + l)...{kr + n-l)V^2^---^k) '
[Z.Z^) lu,r,n ^ ^^^^^ + 1) . . . (A:^ + „ _ 1) 2^ S^...Sr^'
In order to make the upper bound more explicit, we reorganize the n-tuple
(si,...,s^) into those indices t\ < . . . < t(, which appear a certain number of
times ai = (3t. ^2, and those, say tg^i < . . . < t^+nn which appear only once.
We have of course J2 l^i — n — m, and each choice of the t^'s corresponds to
n\/ai\ . . . a£l possibilities for the n-tuple (si, . . . , «„). Therefore we get
^iliiiM<n^f- V -
A trivial comparison series vs. integral yields
^1^1 1
^ ^ a-1 s"-i
S<t<+CXD
and in this way, using successive integrations in ti, ti-i, . . . , we get inductively
^1 1 1
since ctj ^ 2 implies a^-j+i -\- . . . -\- an — i. On the other hand
V 1 ^ 1 1 _ 1 A 1 i"""
. . . ^ m! ^ ..si-.-Sm to! V 2 '■■ fc^
Since partitions ai + . . . + ai = n — m satisfying the additional restriction ai ^ 2
correspond to = ctj — 2 satisfying ^ a'^ = n — m — 2£, their number is equal to
n-m-2£ + £-l^^|^n-m-£-l^ ^
and we infer from this
22
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
where the last term corresponds to the special case £ = 0, m = n. Therefore
si...Sn 2 ^ m\ \ 2 kj\2 k
m=2 V / \ / \
This estimate combined with (2.23, 2.24) implies the upper bound (2.20 b) (the
lower bound 1 being now obvious). The asymptotic estimate (2.20 c) follows
immediately. □
(2.25) Lemma. If A is a hermitian n x n matrix, set lA,q to be equal to 1 if A
has signature in — q, q) and otherwise. Then for all n x n hermitian matrices
A, B we have the estimate
|]lA,qdet^- ]lB,qdet5| ^ ^ ll^llil^l
in— 1— i
ii^ii ii-'-'r
0<i^n-l
where ||^||, ||5|| are the hermitian operator norms of the matrices.
Proof. We first check that the estimate holds true for |detA — deti?|. Let
Ai ^ . . . ^ An be the eigenvalues of A and A'^ ^ ... ^ A^ be the eigenvalues
of B. We have |Aj| ^ ||^||, |A^| ^ ||5|| and the minimax principle implies that
I'^i — '^il ^ 11^ ~ -^11- We then get the desired estimate by writing
det^-detB = Ai...An-Al...A;= J] Ai . . . A,_i(A, - A^A^+i . . . A^.
This already implies (2.25) if ^ or 5 is degenerate. If A and B are non degenerate
we only have to prove the result when one of them (say A) has signature (n — q., q)
and the other one (say B) has a different signature. If we put M{t) = {l — t)A+tB,
the already established estimate for the determinant yields
d
— detM(t)
dt
^ n\\A-B\\ \\M{t)\\ ^n\\A- B\\{{l-t)\\A\\+t\\B
n-l
However, since the signature of M{t) is not the same for t = and t = 1, there
must exist to G ]0, 1[ such that (1 — to)A + toB is degenerate. Our claim follows
by integrating the differential estimate on the smallest such interval [0,to], after
observing that M(0) = A, detM(to) = 0, and that the integral of the right hand
side on [0, 1] is the announced bound. □
(2.26) Lemma. Let Qa be the hermitian quadratic form associated with the
hermitian operator A on C". If fj, is the rotation invariant probability measure
on the unit sphere 5^""-^ of and Aj are the eigenvalues of A, we have
2. Probabilistic estimate of the curvature of A;-jet bundles 23
The norm \\A\\ = max|Aj| satisfies the estimate
I \QA{0?dn{C)^\\A\\\
n J\c\=i
Proof. The first identity as an easy calculation, and the inequalities follow by
computing the eigenvalues of the quadratic form XI -^1 + ( XI -^0 ~ '^^fo^ c > 0.
The lower bound is attained e.g. for Qa(C) — ICiP ~ n(K2p + • • • + ICnP) when
we take in = 1 and c = 1 + -. □
Proof of Proposition 2.19. Take a vector ( G Tx,z, C = SCi^) with ||C||w —
and introduce the trace free sesquilinear quadratic form
where r]ij = X]i<Q,<rCijaa- We consider the corresponding trace free curvature
tensor
(2.27) ®^ ^ ^ X] ^ ® e* (g) 6/3.
As a general matter of notation, we adopt here the convention that the cano-
nical correspondence between hermitian forms and (1, l)-forms is normalized
as ^ttijdzi ® dzj •H- ^'^O'ijdzi A dzj, and we take the liberty of using the
same symbols for both types of objects; we do so especially for gk{z,x,u) and
r){z) = ^ J2Viji^)dzi Adzj = TrQviz). First observe that for all /c-tuples of unit
vectors u = {ui, ...,Uk) e (S'^''"^)'', = {usa)i^a^r, wc havc
f 1 _ - ^
/ ^ (^ijo^pi^)CiCjUsaUsp dfi{u) = V ^V(Q^,^)
J(S2r-l)k .... S i^„/,. -5
where V((52,c) is the variance of Qz,^ on S^'^~^. This is so because we have a
sum over s of independent random variables on (S'^''"^)'^, all of which have zero
mean value. (Lemma 2.26 shows that the variance V(Q) of a trace free hermitian
quadratic form Q{u) = J2i<a<r ^^I'^al'^ ^^^^ sphere 5"^^"^ is equal to
^^^^^^ X] '^a 5 but we only give the formula to fix the ideas). Formula (2.22) yields
XgdukA^) =
r + 1
A,_r' k{kr + iy
Therefore, according to notation (2.15), we obtain the partial variance formula
\gk{z, X, u){C) - g^z, x){C,)\^dvk,r{x)dii{u)
L
(r + 1)
k{kr +
L( E l).(e.,c,c))^
24
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
in which
of \ Oil
l^s^fc 13a. ^ l^s^fe '
^/.(ev'(C,C))' = v(z.^(ey(C,CK^^)/.) = / |(ev(C,CK«)/.|%W-
By integrating over ^ e 5'^""^ C C"^ and applying the left hand inequality in
Lemma 2.26 we infer
where a"t^,ft(0i/) is the standard deviation of Gv^ on 5^"^"^ x S'^'^~^ :
j\(:\u>=i,Hh=i
On the other hand, brutal estimates give the hermitian operator norm estimates
(2-29) \\g,^{z,x)\\^^( J2 -^^s)l\\v{z)L,
(2.30) \\gk{z,x,u)\\^ ^ ( Y -Xsj\\Qv\\uj,h
where
|0vL,/i= sup \{Qv{C,,Qu,u)h\
We use these estimates to evaluate the g-index integrals. The integral associated
with gj^{z,x) is much easier to deal with than gk{z,x,u) since the characteristic
function of the g-index set depends only on z. By Lemma 2.25 we find
\lg^^q{z,x,u) det gk{z,x,u) - 1^,9(2) det gf,{z,x)\
^\\gk{z,x,u) -gk{z,x)\\^ ^ \\gk{z,x,u)\\l\\g^{
0<i<n-l
2. Probabilistic estimate of the curvature of A;-jet bundles
25
The Cauchy-Schwarz inequality combined with (2.28 - 2.30) imphes
1/2
[ \\9k{z,x,u) - gk{z,x)\\lduk,r{x)dii{u)] x
/ ( Y] \\9k{z,x,u)\\l\\gk{z,x)\\Z~^-'] dukA^)dl^i'^)
(iJ,s.^)'"'-"-')""= oe^)
X
by Lemma 2.20 with n replaced by 2n — 2. This is the essential error estimate. As
one can see, the growth of the error mainly depends on the final integral factor,
since the initial multiplicative factor is uniformly bounded over X. In order to get
the principal term, we compute
/ det g^iz, x) di^kA^) = -^'^^^ viz) I ( E ~) ^^^A^)
- ('°«*'"det,(.).
From there we conclude that
/ / lg^A^^^^'^)9k{z,x,uy dvkA^)dlJ'{u)
r"A;" Jx \ k"^ )
The probabilistic estimate 2.19 follows by (2.17). □
(2.31) Remark. If we take care of the precise bounds obtained above, the proof
gives in fact the explicit estimate
[ QU+kr-l _ jn + kr- 1)1 Ik,r,n ( f n j\
Jxoo^L.,,)^^'''*^,..- nm)rikr-l)l [J^''v,,V +ek,r,nJ)
where
J^nil + l/r)y'(j2\Y' j ^c.,.(ev)E^^^'l|Q^II^,'^ll^(^)ll^'"^^"
26
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
and
k X 2n-2 X 1/2
/ r / T- \ ^'^"^ \
1 v^2n-2 2'"(2n-2)! , 1 , , 1A-"^V^^
^ ^ 3 Z^m=2 (2n-2-m)! V"^ 2 " " " A; ^ J
^ + ! + ■■■ + I log^
by the lower and upper bounds of Ik,r,m Ik,r,2n-2 obtained in Lemma 2.20. As
(2n - 2)!/(2n - 2 - m)! ^ (2n - 2)"*, one easily shows that
(2.32) \eu,rA ^ ^^Y^y^' for k ^ e>^-^ .
log A;
Also, we see that the error terms vanish if Qy is identically zero, biit this is of
course a rather unexpected circumstance. In general, ^ince the form is trace
free. Lemma 2.23 applied to the quadratic form u i-> {Qv{Ci O'^i on implies
^{t + l)~^^'^\\Qv\\tjj,h- This yields the simpler bound
8=1 * ^ ''^ i=l
It will be useful to extend the above estimates to the case of sections of
(2.34) L, = O^cc (1) ^ (l (l + 1 + . . . + 1) f)
where F E PicQ(X) is an arbitrary Q-line bundle on X and tt^ : ^ X is
the natural projection. We assume here that F is also equipped with a smooth
hermitian metric hp- In formula (2.20), the renormalized metric r}k{z.,x,u) of Lk
takes the form
(2.35) r]k{z,x,u)= j_ . ^ -—j-gk{z,x,u) + QF,hF{z),
and by the same calculations its expected value is
(2.36) r]{z) := E{r]k{z,;»)) = Q det v *, det h* (z) + QF,hF{z).
Then the variance estimate for r]k — r] is unchanged, and the bounds for r]k are
still valid, since our forms are just shifted by adding the constant smooth term
0F,/iF i^)- The probabilistic estimate 2.18 is therefore still true in exactly the same
2. Probabilistic estimate of the curvature of A;-jet bundles 27
form, provided we use (2.34 - 2.36) instead of the previously defined L/j, rjk and rj.
An application of holomorphic Morse inequalities gives the desired cohomology
estimates for
= hHX?°, O^oc („) » ^jo (l + i + . . . + ,
provided m is sufficiently divisible to give a multiple of F which is a Z-line bundle.
(2.37) Theorem. Let (X, V) be a directed munifold, F ^ X a Q-line bundle,
{V, h) and {F, hp) smooth hermitian structure on V and F respectively. We define
L, = O^cc(l) ® ^*o(I-(l + i + . . . + 1)f),
V = 0dety*,det/i* +0F,h,F-
Then for all q ^ and all m ^ k ^ 1 such that m is sufficiently divisible, we
have
(a) ^ ^J!^il5^(/^^^^^(-l).,"+0((log.)-)),
(c) X(X,«^ Q(Lr )) = . ^ _ ' . {c^{V* F)- + 0((log fe)-^)) .
Green and Griffiths [GG79] already checked the Riemann-Roch calculation
(2.37 c) in the special case V — and F — Ox- Their proof is much simpler since
it relies only on Chern class calculations, but it cannot provide any information
on the individual cohomology groups, except in very special cases where vanishing
theorems can be applied; in fact in dimension 2, the Euler characteristic satisfies
X = — + ^ + h^, hence it is enough to get the vanishing of the top
cohomology group H'^ to infer /i° ^ x ! this works for surfaces by means of a
well-known vanishing theorem of Bogomolov which implies in general
i/"(x,B«Sri«o(^(i + i + ... + i)F)))=o
as soon as Kx ^ F is big and m ^ 1.
In fact, thanks to Bonavero's singular holomorphic Morse inequalities [Bon93],
everything works almost unchanged in the case where V C Tx has singularities
and h is an admissible metric on V (see (0.11)). We only have to find a blow-up
28
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
/U : Xk Xk SO that the resulting pull-backs /U*Lfc and iJ,*V are locally free, and
/U* det h* , fi*'^h,p,e only have divisorial singularities. Then is a (1, l)-current with
logarithmic poles, and we have to deal with smooth metrics on ji* Lf"^ ®0{—mEk)
where Ef~ is a certain effective divisor on Xf~ (which, by our assumption (0.11),
does not project onto X). The cohomology groups involved are then the twisted
cohomology groups
m{x^'',o{Lf^)®dk,m)
where 3k,m = fJ'*{0{—mEk)) is the corresponding multiplier ideal sheaf, and the
Morse integrals need only be evaluated in the complement of the poles, that is on
X{r], q) ^ S where S — Sing(F) U Sing(/i). Since
MMLfn ® 3k,m) C E^^gV* ® (i + i + . . . +
we still get a lower bound for the of the latter sheaf (or for the of the
un- twisted line bundle 0{Lf^) on X^*^). If we assume that Ky <Si F is big, these
considerations also allow us to obtain a strong estimate in terms of the volume,
by using an approximate Zariski decomposition on a suitable blow-up of {X, V) .
The following corollary implies in particular Theorem 0.5.
(2.38) Corollary. If F is an arbitrary Q-line bundle over X , one has
k" (xf «, 0,co („) « 40 (fi (l + i + . . . + i) f) )
when m ':$> k ':$> 1 , in particular there are many sections of the k-jet differentials
of degree m twisted by the appropriate power of F if Ky <^ F is big.
Proof. The volume is computed here as usual, i.e. after performing a suitable
modification ji : X ^ X which converts Ky into an invertible sheaf. There is of
course nothing to prove if Ky ® F is not big, so we can assume \o\{Ky (g) F) > 0.
Let us fix smooth hermitian metrics Hq on Tx and hp on F. They induce a metric
//*(det/iQ ^ ® hp) on ^*{Ky ® F) which, by our definition of Ky, is a smooth
metric. By the result of Fujita [Fuj94] on approximate Zariski decomposition, for
every 5 > 0, one can find a modification ns : Xg — > X dominating /j, such that
fiUKy®F) = 0~^{A + E)
where A and E are Q-divisors, A ample and E effective, with
Vol(yl) = yl^ ^ Vo\{Ky 0F)-6.
If we take a smooth metric Ha with positive definite curvature form ©a,/i^) then
we get a singular hermitian metric hAhE on iJ,g{Ky F) with poles along E, i.e.
2. Probabilistic estimate of the curvature of A;-jet bundles 29
the quotient /iyi/i^//i*(det ^hp) is of the form e~''^ where (p is quasi-psh with
log poles log \ <je\'^ (mod ^^{Xs)) precisely given by the divisor E. We then only
need to take the singular metric h on Tx defined by
h = hQC^'^'^'^'f
(the choice of the factor ^ is there to correct adequately the metric on detV).
By construction h induces an admissible metric on V and the resulting curvature
current 77 = ^KvA^th* + ^F,hF is such that
jl^ri = QA,hA + [-^]) [^] — current of integration on E.
Then the 0-index Morse integral in the complement of the poles is given by
^X(77,0)\S JXg
and (2.38) follows from the fact that 5 can be taken arbitrary small. □
(2.39) Example. In some simple cases, the above estimates can lead to very
explicit results. Take for instance X to be a smooth complete intersection of
multidegree (di, (i2, • • • , (ig) in Pji"*"* and consider the absolute case V = Tx- Then
Kx = Ox{di + ... + ds-n-s-l).
Assume that X is of general type, i.e. '^dj > n + s + 1. Let us equip V — Tx
with the restriction of the Fubini-Study metric h = ©0(1) ; ^ better choice might
be the Kahler-Einstein metric but we want to keep the calculations as elementary
as possible. The standard formula for the curvature tensor of a submanifold gives
where (3 E C°°(A^''^T^ §>> Hom(Tx, 0(dj))) is the second fundamental form. In
other words, by the well known formula for the curvature of projective space, we
have
{eT^,hic,Ou,u) = Id VI' + \{c,u)\^ - \m ■
The curvature p of {Kx,deth*) (i.e. the opposite of the Ricci form Tr QTx,h) is
given by
(2.40) p = - IV QTX,h = IV(/3 A /3*) - (n + l)h ^-{n + l)h.
We take here F = Ox (—a), a e Q+, and we want to determine conditions for the
existence of sections
(2.41) JJ»(x,Bg°Ti»0(-a^(l + l + . .. + !))). m » 1.
30
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
We have to choose Kx ® Ox (—a) ample, i.e. '^dj > n + s + a + 1, and then (by an
appropriate choice of the metric of F = Ox{—(i))j the form t] = Gxx(g)Ox(-a)
be taken to be any positive form cohomologous to (^ dj — {n + s + a + l))h. We
use remark 2.31 and estimate the error terms by considering the Kahler metric
CO = p-\- (n + s + 2)h= (^^dj + l^h.
Inequahty (2.40) shows that co ^ 2h and also that u) ^ Tr(/3 A /3*). From this,
one easily concludes that \\rj\\i_j ^ 1 by an appropriate choice of rj, as well as
||0Tx,ftL,h ^ 1 and \\OTx,h\\iu,h ^ 2. By (2.33), we obtain forn ^ 2
n-l Jx Jx
where Jx'^^ = {^dj + l)"deg(X). On the other hand, the leading term J^r]'^
equals (^dj— n — s — a — l)" deg(X) with deg(X) = di . . .dg. By the bound
(2.32 ) on the error term ek,r,n, we find that the leading coefficient of the growth
of our spaces of sections is strictly controlled below by a multiple of
{j:d,-n-s-a-l) -4.(-)
if k ^ e^^~^. A sufficient condition for the existence of sections in (2.41) is thus
(2.42) ^ exp f 7.38 n^+^Z^ ( ^ , ^^^' + 1 ) .
\ \}_^dj — n — s — a — 1/ /
This is good in view of the fact that we can cover arbitrary smooth complete
intersections of general type. On the other hand, even when the degrees dj tend
to +00, we still get a large lower bound k ~ exp(7.38n'^"'"^/^) on the order of jets,
and this is far from being optimal : Diverio [Div09] has shown e.g. that one can
take k = n for smooth hypersurfaces of high degree. It is however not unlikely
that one could improve estimate (2.42) with more careful choices of u, h. □
3. On the base locus of sections of fe-jet bundles
The final step required for a complete solution of the Green-Griffiths conjecture
would be to calculate the base locus C of the space of sections
iy°(Xp^, OxGG(m) (g) 7rtO(-m5kA)), A ample on X, 5k ^ c^^, c < 1,
k
and to show that = 7r'k{Bk) is a proper algebraic subvariety of X for k large,
under the assumption that Ky is big. This does not look completely hopeless,
since the statistics of curvature in the Morse inequalities do involve currents for
3. On the base locus of sections of A;-jet bundles 31
which the sets of poles depend only on the bigness of Ky and therefore project
onto a proper subvariety S oi X (see the last step of the proof in section 2). It
is not unreasonable to think that a further analysis of the asymptotic behavior of
sections, e.g. through estimates of the Bergman kernel, might lead to such results.
Even if the required property of the base locus cannot be obtained directly, it
would be enough, for a suitable irreducible analytic set Z C X^'^ contained in the
base locus at some stage, to construct non zero sections in
whenever 7Tk{Z) = X, and then to proceed inductively to cut-down the base locus
until one reaches some Z' G Z with iTk{Z') C X. Hence we have to estimate
the cohomology groups and H'^ not just on X^'^, but also on all irreducible
subvarieties Z C X^'^ such that 7rk{Z) = X. We are not able to do this in such a
generality, but our method does provide interesting results in this direction.
(3.1) Theorem. Let {X,V) be a compact directed n-dimensional manifold, let
r = laxikV and F be a holomorphic line bundle on X . Fix an irreducible analytic
set Zk„ C X^^^ or equivalently some C* -invariant set Z^,^ C J'^'^V, and assume
that nkoiZko) = X. For k ^ ko, denote by Zk C Xjf^ the irreducible set cor-
responding to the inverse image of Z'^^ by the canonical morphism J^V — )■ J'^^V .
Let h be an admissible metric onV , hp a metric with analytic singularities on F
and
Lu = 0^cg(1) (8 7r^o(^ (l + ^ + . . . +
V = QKv,deth* + QF,hF, S = Sing(?7).
Then for m ^ k ^ ko and pk = dim Z^ = dim Zj.^ -\- {k — ko)r we have
h\Z,,0{Lf-)\zJ
> !I!!!M^deg^ac/^(Z,)f / rf- 0((log/c)-i)) - o{m^-)
where degj^cG/xl-^fe) = degj^GG/x('^fco) (^)^ relative degree of Zk over X
with respect to the normalized weighted "Kdhler metric" u)a,r,p introduced in (1.10).
We would also get similar upper and lower Morse bounds for the higher
cohomology groups, provided that the sheaves OxGG(m) are twisted by the
appropriate multiplier ideal sheaves 3k,m already described. The main trouble
to proceed further in the analysis of the base locus is that we have to take k ^ ko
and that the 0{...) and o(...) bounds depend on Hence the newer sections
can only be constructed for higher and higher orders k, without any indication
that we can actually terminate the process somewhere, except possibly by some
extremely delicate uniform estimates which seem at present beyond reach.
Proof. The technique is a minor variation of what has been done in section 2,
hence we will only indicate the basic idea. Essentially the A;-jet of / is no longer
32
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
completely random, its projection onto the first ko components (V^/(0))i^j^fcg is
assigned to belong to some given analytic set C J^°V . This means that in the
curvature formula (2.15)
gk{z,X,u) = ^ ^ ^ Cijaf5{z)UsaUs/3dZi AdZj
l^s^fe i,j,a,l3
only the sum '^f.^^g^^j^ is perfectly random. The partial sum ^i<s<fcg remains
bounded, while the harmonic series diverges as log A;. This implies that the
"non randomness" of the initial terms perturbs the estimates merely by bounded
quantities, and in the end, the expected value is still similar to (2.18), i.e.
E{gk{z,.,.)) = i-(^i + l + ... + l + o(i))(eK^,det/.* + eF,ftF).
Once we are there, the calculation of standard deviation and the other estimates
are just routine, and Theorem 3.1 follows again from Proposition 2.13 when we
integrate the Morse integrals over instead of the whole k-jet space
Another possibility to analyze the base locus is to study the restriction maps
(3.2) pk,m{x) : H^{X, E^^^V* ® 0{-mSkA)) ^ {eI^V* ® 0(-m4A))^
at generic points x G X. If Pk,m{x) can be shown to be surjcctivc at a generic
point, then a fortiori the projection = 7rfc(-Bfe) of the base locus does not
contain x and so Yfc is a proper algebraic subvariety of X. Now, proving the
surjectivity of pk,m{x) could be done by proving the vanishing of the group
of our sheaf twisted by the maximal ideal mx,x- We cannot exactly reach such a
precise vanishing result, but Morse inequalities can be used to show that the
groups do not grow too fast.
In fact assume that A is an ample Q-divisor on X which is chosen so small that
Kv®0{—A) is still big. By our estimates, we can then take 5k — ■^(1 + ^ + - ■ ■ + i)-
Pick a very ample divisor G on X and n pencils of sections aj^t £ H^iX., 0(G)),
^ ^ j ^ "n, t E P^, such that the divisors aj^tjiz) = intersect transversally
at isolated points for generic choices of the parameters tj e P^. We select an
admissible metric h on V which provides a strictly positive curvature current on
Kv <S> 0{—A) and multiply it by the additional weight factor (^e'^y/rmSk where
^(^)=iog Yi n
and Tj C P^ are generic finite subsets of given cardinality A^. The multiplier ideal
sheaf of ip is precisely equal to the ideal Je of germs of fimctions vanishing on a
certain 0-dimensional set E — {xi, . . . ,Xs} C X of cardinality s — N'^G'^. Also
the resulting curvature form
1 ^ N
3. On the base locus of sections of A;-jet bundles 33
can be made to be strictly positive as a current provided that ~ cmSk
with c ^ 1 . Then the corresponding multipher ideal sheaf of the induced hermitian
metric on
OxGG (m) (g) TTlOi-mSkA)
k
is the original multiplier sheaf dk,m twisted by tt^J^ above xj, provided that the
Xj lie outside of Sing(F) and outside of the projection of the support V{dk,m)-
Consider the exact sequence
— > OxGG (m) (g) TT^Oi-mSkA) ^ dk,m «) TrpE
— > OxGG (m) (8) 7rlO{-mSkA) ® dk,m
k
Oxoa (m) (8) 7rlO{-m5kA) ® 3k,m ® KC^x/^e) 0.
Its cohomology exact sequence yields an "almost surjective arrow"
H%OxOG{m)®nlO{-m6kA)®dk,m) ^ (^^V^* ® 0(-m5feA))^.,
namely the image contains the kernel of the map
{E^^gV*®0{-m6kA))^, ^H\OxOG{m)(»nlO{-mdkA)®3k,m®K^E).
Now, we have a Morse upper bound
h\Oxoo{m)^nm-mSkA)^3k,m^7rPE) < + fe, _ i); ^! (fc;). ^((^"g ^)"')
since the 1-index integral Jx(^ri i) identically zero. At the same time we have
s = ~ c'm"(log/c)'^/A;",'and it follows that
dim [E^lV* ® 0{-m5kA)) ~ g ,, ""Tw, ^ ~ n'^^TjJ,^ {^ogk^ ^
i<7</ ' {kr-l)\{k\y {kr-l)\{k\Y /c^
By selecting a suitable point Xj and a trivial lower semi-continuity argument we
get the desired almost surjectivity.
(3.3) Corollary. If A is an ample Q-divisor on X such that Ky (8) 0{—A) is big
and Sk = ■^(1 + I + • • • + i)? = ranky, the restriction map
Pk,m{x) : H\X, E^^^V* 0{-m5kA)) ^ {E^^^V* 0{-m5kA))^
has an image of dimension larger than (1 — 0((log /c)^^)) dimE^'^V* at a generic
point X & X for m ^ k ^ 1.
34
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
Such a result puts an upper bound on the vanishing order that a generic
section may have on X^'^ above a generic point of X. Our hope is that one can
then completely "eliminate" the base locus by taking vertical derivatives along
the fibers of j'^V — > X ; those derivations will necessarily have some poles 0{pA)
which we hope to get cancelled by the negative powers 0{—m5kA). This strategy
first devised by [Siu02, Siu04] has indeed been successful in some cases for the
study of generic algebraic degeneracy (e.g. for hypersurfaces of very large degree
in P^"*"^). This would work rather easily if the rough error term 0((log/c)~^)
could be replaced e.g. by 0{m~'^^) in Corollary (3.3), but this is maybe too much
to ask for.
We finally discuss yet another approach. For this we have to introduce
invariant jet differentials along the lines of [Dem95]. In fact, to any directed
manifold (X, V) one can associate its tower of Semple fc-jet spaces, which is a
sequence of directed pairs (X^, Vk) starting with (Xq, Vq) — (X, F), together with
morhisms tt^ : (X^, V^) — )> (X/;_i, Vfc_i). These spaces are constructed inductively
by putting Xfc = P{Vk-i) and Vk = (^fe)-^^^^ (-1)) where
Ox,(-l)c(^fe)*Ffc_iC(^fc)*Tx,_,
is the tautological subbundle (cf. [Dem95]). In the case where V is not a subbundle,
we can first construct the absolute tower {Xk, Vk) by starting from Vq = Tx, and
then take Xk to be the closure in Xk of the A;-step X'^. of the relative tower (X^, V^)
constructed over the dense Zariski open set X' = X ^ Sing(T^). In this way, the
tower {Xk, Vk) is at least birationally well defined - in such a birational context
we can even assume that Xk is smooth after performing a suitable modification at
each stage. Even if we start with V = Tx (or an integrable subbundle V C Tx),
the A;-jet lifting Vk will not be integrable in general, the only exception being when
rankVfe = ranky = 1. Now, if
T^kfi = TTfe O . . . O TTi : Xk Xq — X,
it is shown in [Dem95] that the direct image sheaf
TTkfiOxAm) ■■= Ek,mV*^E^^V*
consists of algebraic differential operators Pifj^^k) which satisfy the invariance
property
when e Gfc is in the group of /c-jets of biholomorphisms ip : (C,0) (C,0).
Since we already assume C* invariance, it is enough to require invariance by
the nilpotent subgroup C Gk of /c-jets tangent to identity. The group
is a semi-direct product of additive groups (C, +) consisting of biholomorphisms
Tj^a t + aP + 0{P'^^), 2 ^ j ^ k, a e C In this tower, the biholomorphisms
3. On the base locus of sections of A;-jet bundles 35
Tfc^o actually generate a normal subgroup of G^, and we have G^/{Tfc,a} ~ '^'k-i-
Now, assume that we have found a section
P e H\X, ^ 0{-m5kA))
for some ample Q-divisor A on A. Then we have an expansion
O^s^m/fe
and the highest non zero term Ps is {r^ a}-invariant of weighted degree m—{k—l)s ;
this comes from the fact that the homothety hx{t) = Xt satisfies
Then it makes sense to look at the action of {rk-i^a} on Pg, and proceeding
inductively we reach a non zero G^-invariant (and thus G^-invariant) polynomial
Q e Ek,m'V* (8) Oi-mSkA))
of degree m' ^ m (and possibly of order k' ^ k but we can still consider it to be
of order k). By raising Q to some power p and using the Q-ampleness of A, we
obtain a genuine integral section
Qp^pim-m')6, ^ H\X,Ek,^m'V* ® 0{-pm'5kA)).
(3.4) Corollary. Let (X, V) he a projective directed manifold such that Ky is big,
and A an ample Q-divisor on X such that Ky ® 0(— A) is still big. Then, if we put
Sk = ■^(1 + I + • • • + i)j = rank V, the space of global invariant jet differentials
H''{X,Ek,mV*(^0{-m6kA))
has (many) non zero sections for m ^ k ^ 1.
If we have a directed projective variety (X, V) with Ky big, we conclude that
there exists k ^ 1 and a proper analytic set Z C X^ such that all entire curves
have the image of their /c-jet f[k]{C) contained in Z. Let Z' be an irreducible
component of Z such that 'Kk,o{Z') = X (if nk,o{Z') C X there is nothing more
to do). Consider the linear subspace V — Tz'^z" H Vk where Z" C Z' is chosen
such that Z' \ Z" is non singular and the intersection Tz'^z" n Vfe is a subbundle
of Tz'^z"- If /[fc](G) is not contained identically in Z" , then the curve g = /[^j
is tangent to {Z',V'). On the other hand, if /[fc](C) C Z" we can replace Z'
by Z" and argue inductively on dimZ'. What we have gained here is that we
have replaced the initial directed space {X, V) with another one {Z' , V) such that
rank V' < rank V, and we can try to argue by induction on r = rank V.
36
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
Observe that the generalized Green-GrifRths conjecture is indeed trivial for
r = 1 (assuming Ky = 0{V*) big) : in fact we get in this case a non zero section
P e H^{X, V*®^ ® for some /c > 1 and so P{f) ■ {f'f must vanish for
every entire curve / : (C,Tc) ^ (X, V). Therefore /(C) dY := {P{z) = 0} C X.
The main difficulty in this inductive approach is that when we start with (X, V)
with Kv big, it seems to be very hard to say anything about Ky' on {Z' ,V').
Especially, the singularities of Z' and V do not seem to be under control. The
only hope would be to have enough control on the sections cutting out Z', and this
requires anyway to understand much more precisely the behavior and vanishing
order of generic sections P e H^{X, Ek^^nV* <8 0{—m5kA)). One could try in
this context to take A to approach the positive part in the Zariski decomposition
of Ky , in such a way that the sections P do not have much space to move around
statistically.
References
[BerlO] Berczi, G.: Thorn polynomials and the Green-GrifRths conjecture. Manuscript
Math. Institute Oxford, May 2010.
[Blo26] Bloch, A.: Sur les systemes de fonctions uniformes satisfaisant a J 'equation d'une
variete algebrique dont I'irregularite depasse la dimension. J. de Math., 5 (1926),
19-66.
[Bog79] Bogomolov, F.A.: Holomorphic tensors and vector bundles on projective varieties.
Math. USSR Izvestija 13/3 (1979), 499-555.
[Bon93] Bonavero, L.: Inegalites de morse holomorphes singulieres. Acad. Sci. Paris Ser. I
Math. 317 (1993), 1163-1166, and: J. Geom. Anal. 8 (1998), 409-425.
[Bro78] Brody, R.: Compact manifolds and hyperbolicity. Trans. Amer. Math. Soc. 235
(1978), 213-219.
[BG77] Brody, R., Green, M.: A family of smooth hyperbolic surfaces in P^. Duke Math.
J. 44 (1977), 873-874.
[Cle86] Clemens, H.: Curves on generic hypersurfaces. Ann. Sci. Ec. Norm. Sup. 19 (1986)
629-636, Erratum: Ann. Sci. Ec. Norm. Sup. 20 (1987) 281.
[CG76] Cowen, M., Griffiths, P.: Holomorphic curves and metrics of negative curvature.
J. Analyse Math. 29 (1976), 93-153.
[Dem85] Demailly, J. -P.: Champs magnetiques et inegalites de Morse pour la d" -cohomologie.
Ann. Inst. Fourier (Grenoble), 35 (1985), 189-229.
[Dem95] Demailly, J. -P.: Algebraic criteria for Kobayashi hyperbolic projective varieties and
jet differentials. AMS Summer School on Algebraic Geometry, Santa Cruz 1995,
Proc. Symposia in Pure Math., ed. by J. Kollar and R. Lazarsfeld, 76p.
[Dem97] Demailly, J. -P.: Varietes hyperboliques et equations differentielles algebriques. Gaz.
Math. 73 (juillet 1997), 3-23.
[DemlOa] Demailly, J.-P. : Holomorphic Morse inequalities and asymptotic cohomology groups:
a tribute to Bernhard Riemann. arXiv: math. CV/1003. 5067.
[DemlOb] Demailly, J.-P.: A converse to the Andreotti-Grauert theorem. Manuscript Inst.
Fourier Grenoble, October 2010.
[DEGOO] Demailly, J.-P., El Goul, J.: Hyperbolicity of generic surfaces of high degree in
projective 3-space. Amer. J. Math. 122 (2000), 515-546.
[Div09] Diverio, S.: Existence of global invariant jet differentials on projective hypersurfaces
of high degree. Math. Ann. 344 (2009), 293-315.
References 37
[DMRIO]
[DTIO]
[Fuj94]
[GG79]
[KobTO]
[Kob76]
[Kob075]
[Lang86]
[Lang87]
[McQ96]
[McQ98]
[McQ99]
[MerOS]
[Mer09]
[MerlO]
[Pau08]
[Rou06a]
[Rou06b]
[Rou07]
[Siu97]
[Siu02]
[Siu04]
[SY96a]
[SY96b]
Diverio, S., Merker, J., Rousseau, E.: Effective algebraic degeneracy. Invent. Math.
180 (2010), 161-223.
Diverio, S., Trapani, S.: A remark on the codimension of the Green- Griffiths locus of
generic projective hypersurfaces of high degree. To appear in J. Reine Angew. Math.
Fujita, T.: Approximating Zariski decomposition of big line bundles. Kodai Math.
J. 17 (1994), 1-3.
Green, M., Griffiths, P.: Two applications of algebraic geometry to entire holomor-
phic mappings. The Chern Symposium 1979, Proc. Internal. Sympos. Berkeley, CA,
1979, Springer- Verlag, New York (1980), 41-74.
Kobayashi, S.: Hyperbolic manifolds and holomorphic mappings. Marcel Dekker,
New York (1970).
Kobayashi, S.: Intrinsic distances, measures and geometric function theory. Bull.
Amer. Math. Soc. 82 (1976), 357-416.
Kobayashi, S., Ochiai, T.: Meromorphic mappings into compact complex spaces of
general type. Invent. Math. 31 (1975), 7-16.
Lang, S.: Hyperbolic and Diophantine analysis. Bull. Amer. Math. Soc. 14 (1986)
159-205.
Lang, S
(1987).
McQuillan, M.: A new proof of the Bloch conjecture
107-117.
Introduction to complex hyperbolic spaces. Springer- Verlag, New York
J. Alg. Geom. 5 (1996),
McQuillan, M.: Diophantine approximation and foliations. Inst. Hautes Etudes Sci.
Publ. Math. 87 (1998), 121-174.
McQuillan, M.: Holomorphic curves on hyperplane sections of 3- folds. Geom. Punct.
Anal. 9 (1999), 370-392.
Merker, J.: Jets de Demailly-Semple dordres 4 et 5 en dimension 2. arXiv:
math.AG/0710.2393, Int. J. Contemp. Math. Sciences 3 (2008) 861-933.
Merker, J.: Low pole order frames on vertical jets of the universal hypersurface.
arXiv: math.AG/0805.3987, Ann. Inst. Fourier (Grenoble) 59 (2009) 1077-1104.
Merker, J.: Complex projective hypersurfaces of general type: toward a conjecture
of Green and Griffiths. Manuscript Ec. Norm. Sup. Paris, May 2010, arXiv:
math.AG/1005.0405.
Paun, M.: Vector fields on the total space of hypersurfaces in the projective space
and hyperbolicity. Math. Ann. 340 (2008) 875-892.
Rousseau, E.: Etude desjets de Demailly-Semple en dimension 3. Ann. Inst. Fourier
(Grenoble) 56 (2006) 397-421.
Rousseau, E.: Equations differentielles sur les hypersurfaces de P^. J. Math. Pures
Appl. 86 (2006) 322-341.
Rousseau, E.: Weak analytic hyperbolicity of generic hypersurfaces of high degree
in P4. Annales Fac. Sci. Toulouse 16 (2007), 369-383.
Siu, Y.T.: A proof of the general schwarz lemma using the logarithmic derivative
lemma. Communication personnelle, avril 1997.
Siu, Y.T.: Some recent transcendental techniques in algebraic and complex geome-
try. Proceedings of the International Congress of Mathematicians, Vol. I (Beijing,
2002), Higher Ed. Press, Beijing, 2002, 439-448.
Siu, Y.T.: Hyperbolicity in complex geometry. The legacy of Niels Henrik Abel,
Springer, Berlin, 2004, 543-566.
Siu, Y.T., Yeung, S.K.: Hyperbolicity of the complement of a generic smooth curve
of high degree in the complex projective plane. Invent. Math. 124 (1996), 573—618.
Siu, Y.T., Yeung, S.K.: Defects for ample divisors of Abelian varieties, Schwarz
lemma and hyperbolic surfaces of low degree. Preprint (prepublication, automne
38
Holomorphic Morse inequalities and the Green-Griffiths-Lang conjecture
1996).
[Tra95] Trapani, S.: Numerical criteria for the positivity of the difference of ample divisors.
Math. Z. 219 (1995), 387-401.
[Voi96] Voisin, C.: On a conjecture of Clemens on rational curves on hypersurfaces. J. Diff.
Geom. 44 (1996), 200-213. Correction: J. Diff. Geom. 49 (1998), 601-611.
(version of November 26, 2010, printed on November 30, 2010)